Toward Correcting InSAR Images for Tropospheric Delay

1
Toward Correcting InSAR Images for Tropospheric
Delay

• A.W. Moore, S.L. Granger, S.E. Owen, F.H. Webb,
  E.J. Fetzer, E.J. Fielding, E.F. Fishbein
• Jet Propulsion Laboratory, California Institute
  of Technology
• C.F. Bjorndahl, J. Lofgren
• Chalmers University of Technology

• Motivation Basic Info InSAR, AIRS, GPS
• Intercomparison of AIRS and GPS
• Stretched Boundary Layer algorithm combining
  ECMWF and DEM
• Selecting time periods conducive to tropospheric
  correction

2
Motivation InSAR Basics
InSAR differencing two radar images to find the
phase difference, and therefore ground
displacement. Spatial density can reach 20m. But
atmospheric differences between the two images
can yield up to 20cm of differential tropospheric
delay in the interferogram, obscuring the true
signal.
Idea produce a tropo correction map to remove
the atmospheric differences from the
interferogram. Bock Williams (1997) suggested
use of GPS limited coverage (few km) has been a
factor. Z. Li colleagues more recently have
investigated use of MERIS and MODIS data
1kmX1km, but not at night and not in presence of
clouds.
3
AIRS Atmospheric Infrared Sounder

• High spectral resolution IR sounder with 2400
  channels.
• Atmospheric profiles at high vertical resolution.
• 2 km vertical resolution for water vapor.
• Nighttime retrievals, in the presence of up to
  70 cloud cover.
• Twice daily retrievals (ascending and descending
  nodes)
• 45km (horizontal) IWV product
• Highest-quality results not available in heavy
  rainfall

4
GPS tropo

• Temporally near-continuous (5min)
• Measures actual delay at L-band
• Spacing few km 10s of km in S. Calif

Zenith delay is formed from all rays passing
through an inverted cone centered on the antenna.
Most variability comes from the lowest 2km of
atmosphere, which implies a cone of R16km at its
top for a cutoff angle of 7 degrees. Most authors
treat the zenith delay as a point solution
directly above the antenna.
32km
2km
5
AIRS/GPS intercomparison methodology

• GPS processed with GIPSY-OASIS II in PPP mode,
  using JPLs Flinn final precise orbit. We used a
  7 degree elevation cutoff and the GMF mapping
  function. Total tropospheric zenith delay and 2
  gradients were estimated as stochastic
  parameters, updated at 5-minute intervals.
• GPS PWV calculated from total delay by method of
  Bevis, et.al. (1992) with surface pressure temp
  from either NCEP(50km) or ECMWF (25km).

AIRS products were generated at the Goddard DAAC
using the v5.0 processing algorithm.
Comparison points limited to measurements
within 25km horizontally, 30m height, and
30minutes.
6
AIRS/GPS PWV Intercomparison
(a)
(b)
GPS, AIRS and ECMWF daily water vapor over Japan,
January 2005 using ECMWF surface pressure to
derive GPS PWV
(a)
(a) GPS approximate ZWD estimated from GEONET GPS
over Japan, January 3, 2005 (b) AIRS PWV over
Japan on January 3 2005

• Japan GEONET with NCEP
• 0.75 correlation coefficient over Jan 05
• Bias evident
• Consistent with previous study Fetzer 2006
• May change with use of absolute antenna
  calibrations

To be completed with data from all seasons, and
other geographic areas.
7
ECMWF DEM GPS

• With Chalmers students J. Lofgren F. Bjorndahl,
  JPL AIRS, GPS InSAR investigators
• Modulates ECMWF (25kmx25km) weather data by 2
  arcsecond (60m) USGS National Elevation Database
  (NED) topographic data in an interpolation
  algorithm to form 60m-resolution TWV maps, in the
  stretched boundary layer approach of E.
  Fishbein
• Will compare the interpolated ZWD with results
  from available Southern California GPS sites
• Will test a differential map as an InSAR
  correction product
• May use GPS data as a correction

Los Angeles area interpolated total PWV map,
2006-07-08 1800
ECMWF
ECMWF el. over this terrain
8
Quiet atmosphere predictor

• GPS troprange differences overlaid on
  interferogram generally correllate
• but are too sparse to sample short-wavelength
  variations between satellite measurements
• Can we select days without high-frequency
  variations?

9
What is noisy atmosphere

• Short-spatial-wavelength activity is associated
  with a passing front, which moves through the
  area
• Other causes include
• Winds
• Boundary layer convection
• Evaporation from local sources of water
• A front passing a single station will leave some
  signature in that stations trop time series
• What can we look at in a single stations time
  series that would indicate a passing front?

10
Quiet atmosphere predictor

• Range of ZWD over the day?
• Max abs(dZWD/dt) over the day?
• Std. Dev. of ZWD over the day?
• Correlated, and vary from day to day
• Look at most different day time series, 11 Jan
  05

11
11 Jan 2005 Trop Behavior
After 3.5 days of continuous rain, the weather
cleared on Jan 11
Jan 10 was La Conchita landslide
12
Jan 19 more favorable atmosphere
Jan 11
Has potential to select days without quickly
varying (temporally) and/or short-wavelength
(spatially) atmosphere features, such that GPS
can be used to run between satellite data
collection more effectively
Jan 19
13
Ahead/summary

• Complete GPS-AIRS comparisons using ECMWF
  pressure temp data
• Evaluate effect of absolute antenna calibrations
  on GPS-AIRS bias
• Evaluate usage of gradients and/or slant delays
  in GPS
• GPSMERISMODISAIRSECMWFDEMInSAR correction
  map???